Feature request: add a true no-prompt dangerous permission level to the permission UI
Summary
Please add a first-class permission level in Claude Code that is explicitly equivalent to --dangerously-skip-permissions, and make it behave as a true no-prompt mode.
Problem
Advanced users can launch Claude Code with --dangerously-skip-permissions, but this behavior is not always exposed clearly as a normal permission level in the UI/settings.
Current friction:
- the dangerous bypass mode is primarily exposed as a startup flag
- users may not have an equivalent clearly selectable permission level in the normal UI/settings flow
- in practice, some users still see prompts or permission behavior that does not feel fully equivalent to a true bypass mode
Requested behavior
Add a clearly labeled permission mode in the standard permission-level system that:
- is selectable from the UI/settings
- is explicitly marked as dangerous
- behaves the same as
--dangerously-skip-permissions - does not ask for permission prompts once enabled
- does not silently downgrade to a less permissive mode mid-session
Why this matters
Some users intentionally run Claude Code in a fully trusted local environment and want the fastest possible workflow. If they explicitly opt into the dangerous mode, the product should respect that choice consistently.
Suggested naming
Something like:
- Dangerous: never ask
- Full access: no prompts
- Bypass permissions (dangerous)
The exact label matters less than making this a clear, first-class, reliable permission level in the product.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗